JMbayes2 0.5.0
Major
jm()
now allows for zero-correlations constraints in the covariance matrix of the random effects. When the mixed models provided in theMixed_objects
argument have been fitted assuming a diagonal matrix for the random effects, this will also be assumed in the joint model (in previous versions, this was ignored). In addition, the new argumentwhich_independent
can be used to specify which longitudinal outcomes are to be assumed independent.jm()
can fit joint models with a combination of interval-censored data and competing risks (e.g., one of the the competing events is interval-censored and the other(s) not).A bug in the
predict()
method causing low AUC values has been corrected.The time-varying ROC and AUC now allow to correct for censoring in the interval
Tstart
toThoriz
using inverse probability of censoring weighting. The default remains model-based weights.
JMbayes2 0.4.1
Major
Portable implementation of parallel computing.
function
area()
has gained the argumenttime_window
that specifies the window of integrating the linear predictor of the corresponding longitudinal outcome.
JMbayes2 0.4.0
Major
Function
tvBrier()
has gained the argumentintegrated
for calculating the integrated Brier score.Function
tvBrier()
has gained the argumenttype_weights
and now also allows to correct for censoring in the intervalTstart
toThoriz
using inverse probability of censoring weighting. The default remains model-based weights.The new function
tvEPCE()
calculates the time-varying expected predictive cross-entropy.This version supports Super Learning for optimizing predictions using cross-validation and a library of joint models. In that regard, the new function
create_folds()
can be used to split a dataset in V-folds of training and test datasets. More information can be found in the corresponding vignette.
JMbayes2 0.3.0
JMbayes2 0.2.0
Major
Dynamic predictions for competing risks data can now be computed. An example is given in the Competing Risks vignette.
Function
jm()
can now fit joint models with a recurrent event process with or without a terminating event. The model accommodates discontinuous risk intervals, and the time can be defined in terms of the gap or calendar timescale. An example is given in the Recurrent Events vignette.
JMbayes2 0.1.7
Major
Added the function
tvBrier()
for calculating time-varying Brier score for fitted joint models. Currently, only right-censored data are supported.Added the functions
calibration_plot()
andcalibration_metrics()
for calculating time-varying calibration plot and calibration metrics for fitted joint models. Currently, only right-censored data are supported.Added new section in the vignette for Dynamic Prediction (available on the website of the package) to showcase the use of the functions mentioned above.
JMbayes2 0.1.6
Major
Added a
predict()
method forjm
objects and a correspondingplot()
for objects of classpredict_jm
for calculating and displaying predictions from joint models. Currently, only standard survival models are covered. Future versions will include predictions from competing risks and multi-state models.Added the functions
tvROC()
andtvAUC()
for calculating time-varying Receiver Operating Characteristic (ROC) curves and the areas under the ROC curves for fitted joint models. Currently, only right-censored data are supported.Added a vignette (available on the website of the package) to explain how (dynamic) predictions are calculated in the package.
JMbayes2 0.1.5
Major
Added two vignettes (available on the website of the package) to showcase joint models with competing risks and joint models with non-Gaussian longitudinal outcomes.
Simplified syntax and additional options for specifying transformation functions of functional forms.
The
slope()
function has gained two new arguments,eps
anddirection
. This allows calculating the difference of the longitudinal profile over a user-specified interval.
JMbayes2 0.1.3
Minor
- Used
parallel::clusterSetRNGStream()
injm_fit()
for distributing the seed in the workers. - Changed the default position of the knots for the B-spline approximation of the log baseline hazard.
JMbayes2 0.1.2
Minor
- Changed calls to
floor()
in the C++ code.